High frequency trading in jeopardy thanks to flash crash

No one could have missed the markets flash drop in May.

On 6th May, US markets fell into a trillion dollar freefall. Accenture’s shares fell from $41 at 2:30 that afternoon to just one cent each by 2:47 whilst Apple’s shares soared 40,000 per cent to reach $100,000 each during the same twenty minute period. Overall, the US stock market lost 9 per cent of value before rebounding just as fast. The result is that the SEC is bringing in rules that will stop a freefall by freezing any equity whose price moves more than 10 per cent in any five minute period. This may deal with the issue but does it deal with the cause?

For over a decade electronic trading has progressed from simple puts and calls to hugely complex structures. A decade ago, FIX Protocol was designed to streamline connectivity between buy and sell side, whilst today it is designed to allow millions of trade offers of indications of interest and order cancellations to take place in real-time, low latency operations.

This point was brought home to me in conversation with Thomson Reuters in the UK, who tell me that, in 2000, they were following 5,000 trading changes per second. This figure had increased to 550,000 changes per second today, and the exchanges expect this to reach over a million by the end of year. In New York, the figures are ten times this level with the US Options Exchange claiming to track four million messages per second across the US trading markets, peaking at ten million per second.

Algorithmic black box structures have stretched the point further, by moving from simple trading of stocks to complex trading strategies that seek arbitrage opportunities at every point of the trade lifecycle. For example, these complex layers of trading strategies have allowed the rise of cross-asset class trading, where a simple trade call for Microsoft stock can be hedged with a link to other tech stocks, such as Apple and IBM, mixed with FX pricing movements and the speed of rises and falls on the Dow and S&P 500, all in real-time. This is why we find the interlinkage and complexities of trading created by such technologies are rapidly becoming more precarious.

A point that was obvious to me back in August 2007 when Goldman Sachs' systems messed up losing thirty percent of the value, or $1.5 billion, of their flagship global equity funds in a week. Goldmans' CFO David Viniar said that this was because they were "seeing things that were 25-standard deviation events, several days in a row." A 25-standard deviation event is meant to only happen once every 100,000 years.

It was obvious that systems seeking arbitrage opportunities could get their models wrong and, since the implosion of liquidity in 2008 when Lehman Brothers failed, we know how fragile these systems can be.

It has not stopped the ongoing march of investment innovation however, with the latest sprinkling of low latency for high frequency trading added to the melting pot.

For example, the CIO of NYSE Euronext said to me recently that the fastest processing in the investment world processes in about 250 microseconds today, or 0.00025 of a second.

In other words, in the time it took you to read the last sentence, about ten seconds, about 40,000 trade movements could have flowed sequentially through the system.

And all of those trades movements need to be time stamped, recorded and indexed for client and regulatory reporting purposes, which is why firms ranging from Barclays to Instinet to Credit Suisse have all received fines from the FSA recently for incorrect time reporting of trades.

And every nanosecond is the difference between getting the trade executed or losing the trade execution opportunity to a competitor. So this march of high frequency complex arbitrage-based automated trading will continue.

Unless regulators stop it that is.

The fact is that the SEC and FSA in the USA and UK respectively are really concerned about such developments, particularly with the issues created by ‘flash orders’, where a sell side firm might change prices in real-time between a client’s order and its execution in order to make a buck at their expense.

Similarly, regulators are fearful of things they do not understand or hear stories about that sound like in appropriate use of information systems, such as ‘dark pools’. Dark pools, where orders are placed unseen by the lit markets, are actually good for liquidity according to institutional investors and trading institutions, but have an issue just due to their title: ‘dark’.

Dark is scary and that’s why rumours of the new European revision of the Markets in Financial Instruments Directive might crack down on dark pool trading structures and worse.

All in all, we are living through a moment in time where the steady march of technology to create innovative trading opportunities are likely to be reined in by regulatory authorities who fear irrational and unexpected market movements in real-time.

The question is: how much of a brake can they put on such innovation?

My answer would be: not much.

After all, most regulations result in unexpected consequences, e.g. MiFID moved a third of all European equities trading from traditional exchanges to high frequency automated trading pools. Do they really want to reverse this trend?

I don’t think so.

So it’s all about checks and balances ... or is that cheques and bank balances?